基于核相关滤波器的SAR车辆目标检测与识别

潘卓; 王宾辉; 高鑫; 王岩飞

doi:10.3724/SP.J.1146.2008.00221

基于核相关滤波器的SAR车辆目标检测与识别

doi: 10.3724/SP.J.1146.2008.00221

计量
- 文章访问数: 3621
- HTML全文浏览量: 92
- PDF下载量: 1618
- 被引次数: 0
出版历程
- 收稿日期: 2008-02-27
- 修回日期: 2008-10-27
- 刊出日期: 2009-05-19

Kernel Correlation Filter for Vehicle Detection and Recognition in SAR Images

摘要

摘要: 针对合成孔径雷达(Synthetic Aperture Radar, SAR)目标检测与识别方法对目标方位角敏感的问题，该文基于相关滤波器理论与核特征分析方法，提出一种对SAR目标方位角具有较强鲁棒性的核相关滤波器。该滤波器使用特征向量降低了对训练图像的依赖性，利用目标在非线性空间的高维特征提高了识别能力，并利用核函数解决了高维矢量的内积计算问题。MSTAR实测SAR图像数据的对比实验结果表明，该文所提出的滤波器在低虚警概率下，能够保持较高的检测概率，并且对目标方位角失真具有较强的容忍性，不需要存储目标模板和估计目标方位角，就能够实现高准确率的目标检测与识别。
- SAR图像;目标检测;目标识别;相关滤波器;核函数
Abstract: SAR target detection and recognition is sensitive to targets azimuth. To solve the problem, based on correlation theory and kernel feature analysis, a kernel correlation filter which is strongly robust to targets azimuth distortion is proposed. The novel filter exploits eigenvectors to reduce the dependence of the training set and extends linear combination of eigenvectors nonlinearly to improve the classification. Moreover, to keep the computation tractable in high dimensional space, the kernel function is employed. Comparative tests using MSTAR database demonstrate the kernel correlation filter performs high detection probability with low false alarm probability and implements target detection and recognition accurately without templates and target poses estimation.

HTML全文

参考文献(1)

Ross T D, Bradley J J, and Hudson L J,et al.. SAR ATR -Sowhats the problem? - An MSTAR perspective. Proceedingsof SPIE-Algorithms for Synthetic Aperture Radar ImageryVI, Orlando, Florida, April 1999, 3721: 662-672.[2]Devore M D and OSullivan J A. Performance complexitystudy of several approaches to automatic target recognitionfrom SAR images[J].IEEE Trans. on Aerospace and ElectronicSystems.2002, 38(2):632-648[3]Shenoy R K. The design and use of unconstrained imagefilters and features for SAR detection and recognition. [Ph.D.dessirtation], Carnegie Mellon University, 2001.[4]Singh R. Advanced correlation filter for multi-class syntheticaperture radar detection and classification. [M.S thesis],Carnegie Mellon University, 2002.[5]Casasent D and Patnaik R. Automated synthesis of distortion-invariant filters: AutoMinace. Proceedings of SPIEIntelligentRobots and Computer Vision XXIV, 2006, Boston,USA, 6384: 638401.[6]Patnaik R and Casasent D. MSTAR object classification andconfuser and clutter rejection using Minace Filters.Proceedings of SPIE-Automatic Target Recognition XVI,Orlando, Florida, April 2006, 6234: 62340S1.[7]Vijaya Kumar B V K. Tutorial survey of composite filterdesigns for optical correlators[J].Applied Optics.1992, 31(23):4773-4801[8]Mahalanobis A, Vijaya Kumar B V K, and Casasent D.Minimum average correlation energy filter[J].Applied Optics.1987, 26(17):3633-3640[9]Rsvichandran G and Casasent D. Minimum noise andcorrelation energy filter[J].Applied Optics.1992, 31(11):1823-1833[10]Isaacs J C, Foo S Y, and Bases A M. Novel kernels and kernelPCA for patten recognitoin. Proceedings of the IEEEInternational sumposium on Computational Intelligence inRobotics and Automation, FL, USA, June, 2007: 438-443.[11]Vijaya Kumar B V K and Xie C Y. Correlation patternrecognition for face recognition[J].Proc. IEEE.2006, 94(11):1963-1975[12] MSTAR Public Release Dataset, website: https:// www.sdms.afrl.af.mil/datasets/mstar/

施引文献

资源附件(0)

访问统计